Elimination of HIV by test and treat

The debate in the public health community about the possibility of eliminating HIV with the so-called ‘test and treat’ strategy was sparked off by an article by Granich et al. [1]. Using a mathematical model, they concluded that testing 90% of a population annually and treating immediately every person found HIV-positive may lead to the elimination of HIV in that population within a time period of around 50 years. In our opinion, some aspects of the work by Granich et al. [1] have not yet received sufficient attention in the debate about HIV elimination. A salient feature of the model is the dependency of the transmission rate (denoted by λ), that is the rate at which an HIV-infected person infects others, on the HIV prevalence in the entire population. This density-dependent transmission parameter supposedly reflects the impact of population heterogeneity in risk behavior on incidence and prevalence. The dependency of an individual's transmission potential on the population prevalence is captured in one parameter (denoted by α), which is not derived from first principles and therefore has no mechanistic interpretation. It is unclear which quantities would have to be measured to provide a quantitative estimate of this parameter. We find that the value of the transmission rate λ has a large influence on the quantitative predictions of the model because it controls the endemic prevalence. In the endemic state, the transmission rate λ is reduced to almost 10% of its baseline value due to density-dependent effects. Any quantitative prediction of the time until elimination, the effort needed to achieve it and the numbers of deaths that can be expected under various scenarios depends on the epidemic dynamics at peak prevalence and therefore on the transmission parameter. Furthermore, the density dependence of transmission and the population growth rate determine the numbers of infections and deaths until elimination. HIV-related deaths may negatively impact on population growth depending on HIV prevalence. Not reaching the elimination goal in an increasing population means increasing numbers of infected individuals even at a constant prevalence. This shows that deriving any statement about the absolute numbers of deaths prevented depends crucially on the assumption about how population heterogeneity impacts on the transmission rate. The question of whether elimination is possible or not at all can be answered independently of the relationship between transmission and prevalence. The possibility of elimination is determined by the effective reproduction number at very low prevalence for which an explicit expression can be derived. We find that for the parameter values used by Granich et al.[1] the elimination threshold is only just reached, meaning that, with slightly less optimistic assumptions on the effectiveness of the intervention, elimination is impossible. In particular, at lower than the assumed 90% coverage, elimination cannot be achieved. The deterministic model used by Granich et al.[1] for projecting the impact of test and treat strategies on prevalence neglects to consider that infectiousness is highly variable during the course of the infection, as is well established through viral load measurement [2] and discordant couples studies [3,4]. A recent analysis estimated the fraction of new infections attributed to early HIV infection in a sub-Saharan African population at around 40% [5]. Similarly high estimates were found in other populations [6–8]. We reimplemented the deterministic model from Granich et al. [1], but modified it by assuming that infectivity (defined as transmission rate times duration of the infectious stage) in the first stage of infection is higher than average and lower in the remaining three stages, while keeping the infectivity over the total infectious period constant. This exercise readily showed that when primary infection (approximated by the first stage of infection) is around 10 times more infectious than the remaining infectious period, elimination cannot be achieved. It should also be taken into account that the recognition of early HIV infection is a major hurdle due to a lack of specific symptoms. Therefore, even if HIV testing could be performed twice a year, detecting early HIV infection in time to reduce the infectiousness of primary infection is hardly possible. Even if the ambitious and possibly unrealistic goals concerning HIV testing uptake, antiretroviral treatment coverage and treatment adherence set by Granich et al.[1] could be met, the elimination goal might be thwarted by a decreasing awareness of the risks of HIV infection resulting in risk compensation. Earlier modeling studies showed that even a moderate increase in risk behavior in combination with increased survival of HIV-infected persons can counteract the positive impact of treatment on prevalence [9,10]. Since the availability of highly active antiretroviral therapy in 1996, a substantial increase in risk behavior and a rising incidence of sexually transmitted infections has occurred among HIV-positive and HIV-negative MSM in industrialized countries [11]. In conclusion, there is no doubt that the test and treat strategy can have a substantial impact on the spread of HIV both in low and high-income countries, but the decision concerning whether to introduce large-scale testing and immediate treatment should not be made with the expectation of reaching HIV elimination within the foreseeable future. Acknowledgements Conflicts of interest There are no conflicts of interest.